import os
import matplotlib.pyplot as plt
import clip
import torch
from torchvision.datasets import CIFAR100
#from torchvision.datasets import CocoDetection
import numpy as np
import PIL
%matplotlib inline
#Had OpenMPI issue
os.environ['KMP_DUPLICATE_LIB_OK']='True'
%load_ext autoreload
%autoreload 2
# Load constellations
from dataUtils import *
pictures = loadConstellations(pictureTypes=["original", "final_easy", "outline"])
### Visualize selected image and its original
imageIdx = 24
objectId = list(pictures.keys())[imageIdx]
constellationImage = pictures[objectId]["final_easy"][0]
outlineImage = pictures[objectId]["outline"][0]
originalImage = pictures[objectId]["original"][0]
fig, (ax1, ax2, ax3) = plt.subplots(nrows=1, ncols=3, figsize=(10, 8))
ax1.imshow(constellationImage)
ax2.imshow(outlineImage)
ax3.imshow(originalImage)
<matplotlib.image.AxesImage at 0x1505c0a300a0>
# Load the model
device = "cuda" if torch.cuda.is_available() else "cpu"
#Different architectures, choose one
model, preprocess = clip.load('ViT-B/32', device) #Faster
#model, preprocess = clip.load('RN50x4', device) #Better but takes more time
# Download the dataset (not necessary, as only interested in classes)
#cifar100 = CIFAR100(root=os.path.expanduser("~/.cache"), download=True, train=False)
#Coco = CocoDetection(root=os.path.expanduser("~/.cache"), download=True, train=False)
#Available classifications from different datasets
#combinedClasses.txt - CoCo + CIFAR100 (around 160 classes, removed duplicates)
#combinedClassesAdditional.txt -CoCo + CIFAR100 + classes about pictures which constellations are in set but
#not in either CoCo or CIFAR100 classes (around 180 classes, removed duplicates)
#cocoClasses.txt - CoCo (80 classes)
#cifarClasses.txt - Cifar100 (100 classes)
#imagenet_classes.txt - Imagenet classes (around 1000 classes)
chosenDataset = "combinedClassesAdditional.txt"
# Read the categories
with open(os.path.join("classes",chosenDataset), "r") as f:
categories = [s.strip() for s in f.readlines()]
# Prepare the inputs
#image, class_id = cifar100[3637]
n = len(pictures.keys())
for i in range(n):
objectId = list(pictures.keys())[i]
constellationImage = pictures[objectId]["final_easy"][-1]
outlineImage = pictures[objectId]["outline"][0]
originalImage = pictures[objectId]["original"][0]
print(objectId)
#print(labels[i])
#print(labels[i] in categories)
fig, axes = plt.subplots(nrows=1, ncols=3, figsize=(10, 8))
axes[0].imshow(constellationImage)
axes[1].imshow(outlineImage)
axes[2].imshow(originalImage)
plt.show()
for image_type in ["final_easy","outline", "original"]:
print(image_type)
# for image_type in ["outline", "final_easy"]:
image = PIL.Image.fromarray(pictures[objectId][image_type][0])
image_input = preprocess(image).unsqueeze(0).to(device)
text_inputs = torch.cat([clip.tokenize(f"a photo of a {c}") for c in categories]).to(device)
# Calculate features
with torch.no_grad():
image_features = model.encode_image(image_input)
text_features = model.encode_text(text_inputs)
# Pick the top 5 most similar labels for the image
image_features /= image_features.norm(dim=-1, keepdim=True)
text_features /= text_features.norm(dim=-1, keepdim=True)
similarity = (100.0 * image_features @ text_features.T).softmax(dim=-1)
values, indices = similarity[0].topk(5)
# Print the result
print("\nTop predictions for", image_type, ":\n")
for value, index in zip(values, indices):
print(f"{categories[index]:>16s}: {100 * value.item():.2f}%")
object_413
final_easy
Top predictions for final_easy :
plain: 5.57%
giraffe: 2.75%
sea: 2.74%
ray: 2.52%
plate: 2.51%
outline
Top predictions for outline :
shirt: 70.91%
person: 1.83%
plain: 1.46%
lawn_mower: 1.25%
cattle: 1.17%
original
Top predictions for original :
shirt: 99.19%
plain: 0.07%
baby: 0.05%
tie: 0.03%
boy: 0.03%
object_451
final_easy
Top predictions for final_easy :
plain: 7.78%
sea: 4.76%
cell phone: 3.35%
plate: 2.98%
kite: 1.95%
outline
Top predictions for outline :
cattle: 76.21%
cow: 17.47%
sheep: 2.20%
bear: 0.84%
lawn_mower: 0.65%
original
Top predictions for original :
cow: 58.27%
cattle: 40.77%
horse: 0.37%
lawn_mower: 0.26%
tractor: 0.07%
object_244
final_easy
Top predictions for final_easy :
plain: 8.69%
sea: 5.78%
cell phone: 3.03%
plate: 2.53%
ray: 2.39%
outline
Top predictions for outline :
motorcycle: 86.54%
lawn_mower: 6.74%
bicycle: 1.03%
car: 0.63%
mouse: 0.42%
original
Top predictions for original :
motorcycle: 94.21%
orange: 1.24%
bicycle: 0.93%
rocket: 0.51%
carrot: 0.24%
object_338
final_easy
Top predictions for final_easy :
plain: 7.17%
plate: 3.20%
ray: 2.87%
sports ball: 2.77%
willow_tree: 2.57%
outline
Top predictions for outline :
chair: 99.02%
couch: 0.21%
table: 0.15%
dining table: 0.12%
shirt: 0.06%
original
Top predictions for original :
chair: 98.95%
table: 0.31%
dining table: 0.30%
bench: 0.12%
plain: 0.05%
object_196
final_easy
Top predictions for final_easy :
plain: 8.42%
sea: 5.67%
plate: 3.14%
kite: 2.62%
cell phone: 2.60%
outline
Top predictions for outline :
flatfish: 22.15%
ray: 21.25%
whale: 10.03%
trout: 4.86%
mouse: 2.79%
original
Top predictions for original :
bird: 90.86%
baby: 1.71%
boy: 1.08%
sweet_pepper: 0.64%
girl: 0.47%
object_345
final_easy
Top predictions for final_easy :
plain: 12.00%
sea: 6.10%
cell phone: 3.32%
mountain: 2.19%
giraffe: 2.17%
outline
Top predictions for outline :
pear: 10.29%
baby: 4.90%
duck: 4.56%
person: 3.43%
beaker: 3.16%
original
Top predictions for original :
toilet: 96.84%
chair: 1.78%
sink: 0.25%
plain: 0.12%
bowl: 0.11%
object_452
final_easy
Top predictions for final_easy :
plain: 6.87%
plate: 4.35%
ray: 3.85%
kite: 3.25%
frisbee: 2.75%
outline
Top predictions for outline :
pistol: 34.63%
wrench: 5.27%
kite: 5.07%
ray: 5.04%
whale: 4.35%
original
Top predictions for original :
scissors: 96.84%
knife: 2.18%
pen: 0.46%
wrench: 0.21%
fork: 0.07%
object_385
final_easy
Top predictions for final_easy :
sea: 8.36%
plain: 5.84%
cell phone: 3.14%
kite: 3.10%
plate: 2.75%
outline
Top predictions for outline :
bear: 28.88%
beaver: 13.10%
sheep: 9.59%
elephant: 7.32%
mouse: 5.55%
original
Top predictions for original :
hippo: 99.73%
cow: 0.05%
elephant: 0.03%
bear: 0.02%
horse: 0.02%
object_339
final_easy
Top predictions for final_easy :
plain: 6.64%
sea: 5.13%
plate: 3.60%
cell phone: 3.34%
kite: 3.08%
outline
Top predictions for outline :
car: 87.14%
pickup_truck: 1.80%
shirt: 0.84%
truck: 0.67%
beetle: 0.49%
original
Top predictions for original :
car: 67.87%
pickup_truck: 2.73%
baby: 2.46%
truck: 2.21%
dinosaur: 1.77%
object_322
final_easy
Top predictions for final_easy :
plain: 6.27%
sea: 3.02%
whale: 2.71%
ray: 2.55%
cell phone: 2.42%
outline
Top predictions for outline :
cat: 52.50%
dog: 3.08%
bear: 2.47%
tiger: 2.14%
sweet_pepper: 1.95%
original
Top predictions for original :
cat: 56.39%
sweet_pepper: 29.83%
boy: 2.51%
mouse: 1.16%
baby: 1.13%
object_453
final_easy
Top predictions for final_easy :
plain: 7.64%
sea: 4.98%
plate: 3.51%
cell phone: 3.18%
frisbee: 2.88%
outline
Top predictions for outline :
lawn_mower: 22.05%
elephant: 13.40%
cattle: 8.32%
cow: 4.78%
sheep: 4.73%
original
Top predictions for original :
cannon: 99.71%
pistol: 0.06%
tank: 0.05%
hair drier: 0.03%
castle: 0.02%
object_369
final_easy
Top predictions for final_easy :
sea: 7.82%
plain: 6.83%
cell phone: 3.43%
plate: 2.89%
cup: 2.34%
outline
Top predictions for outline :
pistol: 90.71%
wrench: 0.84%
lawn_mower: 0.77%
remote: 0.71%
telephone: 0.61%
original
Top predictions for original :
pistol: 98.41%
cake: 0.49%
backpack: 0.14%
book: 0.08%
plate: 0.07%
object_468
final_easy
Top predictions for final_easy :
plain: 9.63%
sea: 3.42%
cell phone: 3.38%
plate: 2.62%
mountain: 2.25%
outline
Top predictions for outline :
aquarium fish: 44.84%
shark: 17.98%
flatfish: 8.24%
ray: 7.70%
whale: 6.59%
original
Top predictions for original :
goldfish: 90.68%
aquarium fish: 9.00%
trout: 0.15%
orange: 0.02%
sweet_pepper: 0.02%
object_448
final_easy
Top predictions for final_easy :
plain: 11.86%
sea: 10.10%
cell phone: 3.17%
plate: 2.13%
mountain: 2.12%
outline
Top predictions for outline :
horse: 15.86%
dog: 10.89%
seahorse: 9.33%
bear: 6.50%
cattle: 5.39%
original
Top predictions for original :
seahorse: 99.96%
aquarium fish: 0.01%
caterpillar: 0.01%
goldfish: 0.01%
worm: 0.00%
object_266
final_easy
Top predictions for final_easy :
plain: 10.58%
sea: 9.95%
cell phone: 2.89%
cup: 2.45%
plate: 2.17%
outline
Top predictions for outline :
duck: 64.92%
penguin: 19.06%
bird: 8.85%
snowman: 0.59%
plain: 0.44%
original
Top predictions for original :
duck: 98.21%
bird: 0.99%
boy: 0.07%
person: 0.07%
man: 0.07%
object_230
final_easy
Top predictions for final_easy :
sea: 8.82%
plain: 7.83%
cell phone: 4.11%
plate: 2.38%
ray: 2.17%
outline
Top predictions for outline :
chair: 98.16%
table: 0.49%
dining table: 0.48%
shirt: 0.14%
bench: 0.05%
original
Top predictions for original :
chair: 97.36%
table: 0.71%
dining table: 0.47%
fire extinguisher: 0.34%
bench: 0.22%
object_263
final_easy
Top predictions for final_easy :
plain: 6.38%
sports ball: 4.04%
cell phone: 4.04%
pine_tree: 3.60%
willow_tree: 3.08%
outline
Top predictions for outline :
chair: 99.32%
couch: 0.15%
television: 0.07%
vase: 0.05%
table: 0.04%
original
Top predictions for original :
chair: 98.79%
table: 0.63%
dining table: 0.21%
plain: 0.03%
cup: 0.02%
object_471
final_easy
Top predictions for final_easy :
plain: 7.04%
cell phone: 3.78%
sea: 3.56%
plate: 2.69%
ray: 2.17%
outline
Top predictions for outline :
fork: 8.19%
pen: 4.64%
skis: 4.51%
toothbrush: 4.16%
tennis racket: 4.08%
original
Top predictions for original :
pen: 99.65%
knife: 0.05%
plain: 0.04%
baseball bat: 0.04%
table: 0.02%
object_473
final_easy
Top predictions for final_easy :
plain: 9.14%
sea: 4.90%
cell phone: 4.08%
plate: 2.36%
ray: 2.06%
outline
Top predictions for outline :
bottle: 48.05%
vase: 42.12%
wine glass: 3.10%
beaker: 1.73%
flask: 1.08%
original
Top predictions for original :
vase: 96.05%
bottle: 1.36%
beaker: 1.05%
wine glass: 0.89%
bowl: 0.17%
object_457
final_easy
Top predictions for final_easy :
plain: 8.85%
sea: 5.98%
cell phone: 3.76%
plate: 2.86%
pine_tree: 2.44%
outline
Top predictions for outline :
lawn_mower: 39.14%
motorcycle: 14.91%
mouse: 4.53%
pickup_truck: 4.12%
tank: 3.61%
original
Top predictions for original :
tank: 98.10%
tiger: 0.78%
cannon: 0.27%
lawn_mower: 0.13%
tractor: 0.12%
object_412
final_easy
Top predictions for final_easy :
plain: 9.46%
sea: 4.89%
pine_tree: 3.22%
cell phone: 3.03%
plate: 2.53%
outline
Top predictions for outline :
turtle: 10.56%
shrew: 9.11%
mouse: 8.45%
snail: 6.66%
dinosaur: 5.86%
original
Top predictions for original :
snail: 98.49%
worm: 0.77%
caterpillar: 0.15%
snake: 0.13%
cockroach: 0.07%
object_411
final_easy
Top predictions for final_easy :
plain: 7.48%
sea: 4.56%
cell phone: 3.35%
plate: 2.96%
sports ball: 2.49%
outline
Top predictions for outline :
ray: 16.86%
kite: 7.57%
flatfish: 6.71%
whale: 4.95%
hand: 4.38%
original
Top predictions for original :
scissors: 87.02%
worm: 2.41%
knife: 2.10%
wrench: 1.47%
spoon: 1.29%
object_409
final_easy
Top predictions for final_easy :
plain: 10.91%
sea: 3.59%
plate: 3.52%
cell phone: 3.22%
kite: 2.65%
outline
Top predictions for outline :
snowman: 98.92%
penguin: 0.26%
teddy bear: 0.12%
bear: 0.09%
cat: 0.06%
original
Top predictions for original :
snowman: 99.81%
fire hydrant: 0.03%
skis: 0.03%
snowboard: 0.01%
person: 0.01%
object_351
final_easy
Top predictions for final_easy :
plain: 7.16%
sea: 5.95%
plate: 3.52%
sports ball: 3.15%
cell phone: 3.09%
outline
Top predictions for outline :
spoon: 77.01%
fork: 5.00%
bowl: 4.38%
flatfish: 1.21%
plate: 0.92%
original
Top predictions for original :
tennis racket: 99.80%
sports ball: 0.08%
baseball bat: 0.03%
frisbee: 0.01%
cockroach: 0.01%
object_432
final_easy
Top predictions for final_easy :
plain: 9.05%
sea: 4.46%
cell phone: 4.00%
giraffe: 2.36%
plate: 2.17%
outline
Top predictions for outline :
bottle: 69.53%
lamp: 7.94%
beaker: 5.22%
fire extinguisher: 2.38%
vase: 2.29%
original
Top predictions for original :
beaker: 72.67%
bottle: 20.34%
wine glass: 3.59%
vase: 0.95%
flask: 0.80%
object_440
final_easy
Top predictions for final_easy :
plain: 9.61%
sea: 6.16%
cell phone: 3.46%
plate: 3.09%
giraffe: 2.29%
outline
Top predictions for outline :
hand: 36.10%
wrench: 8.19%
baseball glove: 6.49%
pistol: 5.07%
kite: 3.64%
original
Top predictions for original :
scissors: 68.17%
wrench: 22.59%
knife: 1.60%
spoon: 1.06%
fork: 1.04%
object_341
final_easy
Top predictions for final_easy :
plain: 8.38%
cell phone: 5.01%
sea: 4.19%
plate: 3.44%
sports ball: 3.01%
outline
Top predictions for outline :
giraffe: 41.06%
camel: 21.27%
dog: 8.90%
dinosaur: 4.87%
horse: 3.80%
original
Top predictions for original :
giraffe: 99.41%
zebra: 0.18%
camel: 0.11%
leopard: 0.07%
elephant: 0.04%
object_243
final_easy
Top predictions for final_easy :
plain: 9.48%
plate: 2.96%
cell phone: 2.90%
sea: 2.74%
kite: 2.10%
outline
Top predictions for outline :
sports ball: 5.03%
spider: 4.72%
flower: 4.47%
plate: 2.34%
frisbee: 2.25%
original
Top predictions for original :
flower: 70.33%
plain: 10.97%
lawn_mower: 3.32%
girl: 1.77%
sunflower: 1.16%
object_442
final_easy
Top predictions for final_easy :
plain: 9.58%
sea: 4.74%
cell phone: 3.41%
pine_tree: 2.85%
plate: 2.48%
outline
Top predictions for outline :
teddy bear: 22.31%
person: 12.17%
boy: 10.94%
bear: 4.65%
snowman: 2.58%
original
Top predictions for original :
baby: 54.11%
boy: 32.85%
hippo: 1.50%
goldfish: 1.01%
person: 0.94%
object_201
final_easy
Top predictions for final_easy :
plain: 9.45%
sea: 5.69%
cell phone: 3.13%
plate: 2.41%
mountain: 2.30%
outline
Top predictions for outline :
penguin: 97.70%
whale: 0.44%
seal: 0.14%
bird: 0.08%
plain: 0.08%
original
Top predictions for original :
penguin: 98.03%
aquarium fish: 0.49%
duck: 0.23%
bird: 0.22%
seal: 0.17%
object_419
final_easy
Top predictions for final_easy :
plain: 8.33%
sea: 5.44%
cell phone: 3.53%
plate: 3.11%
kite: 2.34%
outline
Top predictions for outline :
bicycle: 78.28%
motorcycle: 8.70%
lawn_mower: 2.65%
shirt: 1.38%
skateboard: 0.59%
original
Top predictions for original :
bicycle: 36.17%
baby: 13.30%
backpack: 12.58%
umbrella: 9.98%
kite: 3.99%
object_324
final_easy
Top predictions for final_easy :
sea: 12.73%
plain: 8.57%
cell phone: 3.61%
ray: 2.43%
plate: 2.08%
outline
Top predictions for outline :
flower: 36.62%
sunflower: 21.24%
poppy: 11.75%
rose: 6.79%
clock: 2.05%
original
Top predictions for original :
flower: 82.64%
sunflower: 2.99%
plain: 2.19%
bee: 1.88%
potted plant: 1.26%
object_206
final_easy
Top predictions for final_easy :
plain: 8.54%
sea: 7.38%
cell phone: 3.40%
plate: 2.58%
ray: 1.98%
outline
Top predictions for outline :
dog: 21.70%
wolf: 21.04%
bear: 12.62%
cattle: 7.41%
cow: 4.74%
original
Top predictions for original :
wolf: 98.92%
dog: 0.53%
fox: 0.28%
lion: 0.02%
woman: 0.02%
object_478
final_easy
Top predictions for final_easy :
plain: 8.43%
plate: 2.98%
cell phone: 2.95%
sports ball: 2.24%
kite: 2.24%
outline
Top predictions for outline :
bottle: 88.35%
fire extinguisher: 1.16%
flask: 1.11%
can: 0.71%
wine glass: 0.71%
original
Top predictions for original :
bottle: 88.12%
wine glass: 3.82%
beaker: 0.63%
table: 0.39%
cup: 0.34%
object_449
final_easy
Top predictions for final_easy :
plain: 8.53%
sea: 5.03%
cell phone: 3.36%
plate: 2.93%
giraffe: 1.95%
outline
Top predictions for outline :
pear: 32.99%
beaker: 9.54%
wine glass: 3.67%
vase: 3.41%
plain: 2.74%
original
Top predictions for original :
pear: 99.87%
apple: 0.03%
orange: 0.03%
sweet_pepper: 0.01%
spoon: 0.01%
object_234
final_easy
Top predictions for final_easy :
plain: 9.33%
sea: 4.56%
cell phone: 3.38%
giraffe: 3.26%
ray: 2.60%
outline
Top predictions for outline :
penguin: 19.96%
bear: 7.81%
beaver: 7.23%
chimpanzee: 5.78%
skunk: 5.61%
original
Top predictions for original :
penguin: 99.23%
seal: 0.14%
bird: 0.07%
snowman: 0.05%
plain: 0.04%
object_256
final_easy
Top predictions for final_easy :
plain: 7.54%
sea: 5.82%
cell phone: 2.80%
plate: 2.73%
frisbee: 1.91%
outline
Top predictions for outline :
bottle: 79.98%
fire extinguisher: 4.13%
can: 3.54%
vase: 2.69%
flask: 1.96%
original
Top predictions for original :
fire extinguisher: 99.72%
fire hydrant: 0.23%
can: 0.02%
bottle: 0.02%
flask: 0.00%
object_439
final_easy
Top predictions for final_easy :
sea: 7.45%
plain: 7.17%
kite: 3.68%
plate: 2.81%
cell phone: 2.19%
outline
Top predictions for outline :
hand: 68.21%
pistol: 3.40%
person: 2.99%
baseball glove: 2.62%
beaker: 2.34%
original
Top predictions for original :
hand: 94.87%
woman: 0.81%
person: 0.73%
girl: 0.55%
plain: 0.34%
object_410
final_easy
Top predictions for final_easy :
plain: 8.57%
sea: 5.26%
plate: 2.64%
kite: 2.32%
cell phone: 2.15%
outline
Top predictions for outline :
kite: 9.03%
fork: 7.50%
tie: 6.84%
umbrella: 6.55%
wrench: 6.48%
original
Top predictions for original :
kite: 99.79%
plain: 0.05%
umbrella: 0.03%
ray: 0.02%
rocket: 0.02%
object_229
final_easy
Top predictions for final_easy :
plain: 9.44%
sea: 6.99%
cell phone: 3.21%
mountain: 2.42%
giraffe: 2.39%
outline
Top predictions for outline :
sunglasses: 96.10%
plain: 0.27%
shirt: 0.22%
snake: 0.17%
tie: 0.15%
original
Top predictions for original :
sunglasses: 97.65%
plain: 0.93%
tie: 0.20%
person: 0.09%
shirt: 0.06%
object_248
final_easy
Top predictions for final_easy :
plain: 7.37%
sea: 4.47%
cell phone: 3.07%
kite: 2.84%
plate: 2.76%
outline
Top predictions for outline :
bear: 21.96%
skunk: 15.89%
leopard: 7.52%
mouse: 5.79%
cat: 4.79%
original
Top predictions for original :
squirrel: 99.06%
sweet_pepper: 0.18%
skunk: 0.08%
mouse: 0.08%
baby: 0.06%
object_391
final_easy
Top predictions for final_easy :
plain: 8.69%
sea: 8.55%
cell phone: 3.59%
plate: 2.35%
cup: 2.11%
outline
Top predictions for outline :
boy: 48.06%
person: 26.25%
man: 4.03%
plain: 1.80%
baby: 1.20%
original
Top predictions for original :
boy: 92.27%
baby: 2.90%
person: 2.13%
man: 0.50%
dinosaur: 0.15%
object_447
final_easy
Top predictions for final_easy :
plain: 11.46%
sea: 4.23%
cell phone: 3.65%
plate: 2.98%
mountain: 2.15%
outline
Top predictions for outline :
worm: 10.66%
whale: 8.50%
snake: 3.94%
scissors: 3.37%
sunglasses: 3.29%
original
Top predictions for original :
banana: 98.87%
plain: 0.15%
spoon: 0.14%
apple: 0.06%
cake: 0.05%
object_463
final_easy
Top predictions for final_easy :
plain: 7.86%
sea: 4.30%
plate: 2.94%
ray: 2.88%
kite: 2.86%
outline
Top predictions for outline :
spoon: 21.23%
tennis racket: 13.83%
fork: 7.21%
toothbrush: 6.20%
scissors: 5.98%
original
Top predictions for original :
spoon: 43.86%
baseball bat: 34.17%
tennis racket: 4.92%
skis: 3.80%
knife: 3.11%
object_211
final_easy
Top predictions for final_easy :
plain: 6.25%
sea: 6.19%
kite: 3.50%
cell phone: 3.20%
ray: 3.06%
outline
Top predictions for outline :
pistol: 98.53%
pickup_truck: 0.13%
wrench: 0.11%
shirt: 0.11%
teddy bear: 0.05%
original
Top predictions for original :
pistol: 98.43%
teddy bear: 0.25%
baby: 0.15%
backpack: 0.10%
mouse: 0.07%
object_367
final_easy
Top predictions for final_easy :
plain: 9.69%
sea: 6.83%
cell phone: 3.79%
ray: 2.06%
whale: 2.02%
outline
Top predictions for outline :
hair drier: 9.57%
elephant: 7.40%
telephone: 4.52%
camel: 3.76%
dinosaur: 3.37%
original
Top predictions for original :
hair drier: 99.92%
toothbrush: 0.01%
remote: 0.01%
skis: 0.01%
toaster: 0.00%
object_455
final_easy
Top predictions for final_easy :
sea: 7.46%
plain: 5.91%
ray: 3.43%
kite: 3.39%
plate: 2.73%
outline
Top predictions for outline :
wrench: 51.08%
pistol: 31.79%
hair drier: 1.79%
knife: 1.26%
pen: 1.10%
original
Top predictions for original :
wrench: 99.50%
scissors: 0.37%
knife: 0.07%
spoon: 0.02%
fork: 0.01%
object_225
final_easy
Top predictions for final_easy :
sea: 8.37%
plain: 7.61%
cell phone: 4.36%
plate: 3.61%
kite: 2.50%
outline
Top predictions for outline :
elephant: 43.82%
cow: 17.08%
cattle: 13.67%
bear: 12.04%
hippo: 4.73%
original
Top predictions for original :
elephant: 98.65%
camel: 0.16%
giraffe: 0.13%
baby: 0.11%
lion: 0.06%